A consultant is scoping predictive lead scoring for a two-year-old specialty retailer whose Dynamics 365 Sales environment holds only 18 closed leads total, most of them closed within the last six weeks after a recent one-time sales push. What should the consultant tell the client about enabling predictive lead scoring right now?
Select an answer to reveal the explanation.
Short Explanation
Think of predictive scoring like a new employee learning to spot good prospects by watching hundreds of past deals play out, both wins and losses, over a stretch of time. Eighteen closed records bunched into six weeks is like asking that employee to form a judgment after watching one short sales sprint, they will latch onto whatever quirks happened to show up in that narrow window rather than real patterns. The instinct to just flip the feature on because the tables exist, or to only worry about one of the two tables, or to hand-enter some starting scores, all miss the same point: the model earns its accuracy by studying a broad, varied history of outcomes, not by being switched on or fed a shortcut. The right move here is patience, let the client keep working leads and opportunities normally, and revisit enabling the feature once there is a deeper and more time-varied body of closed records for it to learn from.
Full Explanation
The correct answer is A. Predictive lead scoring learns to distinguish likely winners from likely losers by studying a body of closed records, and a model trained on 18 records clustered into a six-week window has neither enough volume nor enough time span to capture real patterns rather than noise from one campaign. The consultant should recommend waiting until closed leads and opportunities accumulate across a longer, more representative period. Option B is incorrect because the model's value comes specifically from learning against historical closed outcomes, not merely from the Leads table existing. Option C is incorrect because it understates the requirement on both sides: opportunities also need a meaningful body of closed history, and skipping leads entirely does not solve the underlying data-volume problem the client actually has. Option D is incorrect because there is no mechanism for manually assigning scores to bootstrap the predictive model; scores are computed by the trained model itself, and manually entering values would not train anything or persist once real scoring runs.